Abstract
The prognostic value of nuclear features based on tumor-associated collagen signatures (TCMF2) is still unclear. In this paper, we extracted and quantified the TCMF2 from 941 invasive breast cancer patients in H&E images. The least absolute shrinkage and selection operator regression were used to build a TCMF2-score. The univariate and multivariate Cox proportional hazards regression analyses showed that the TCMF2-score is an independent prognostic factor with an advantage in the prognosis of early-stage invasive breast cancer. When the TCMF2, the microscopic features of TACS-based collagen (TCMF1) and the tumor-associated collagen signatures (TACS) were combined, they showed better accuracy in patient stratification than the clinical model (CLI) or the model based on TACS + TCMF1. Our results identify that TCMF2 improves the performance of the TACS-based prediction model, and the TACS-based full model (TACS + TCMF1 + TCMF2) may help us stratify patients more accurately and provide more appropriate adjuvant therapy.